FRET is a physicochemical phenomenon which only occurs when two fluorophores are in sufficient proximity (<10 nm) of each other, and the emission spectrum of the donor overlaps the excitation spectrum of the acceptor [57]. In the AFP-based FRET strategy, CFP and YFP mutants have been favorably utilized as a FRET donor and an acceptor, respectively [58]. Engineering with two AFPs in combination with the receptor protein affords a sensor protein that responds to dynamic fluctuation of intracellular ligand concentration by a ratiometric fluorescence change. The feasibility of this strategy strongly depends on the magnitude of the structural change of the receptor. In the case of the receptor with a large structural change upon binding the substrate, this strategy would be the most straightforward way to integrate the signal transduction function into the receptor of interest (Figure 1a), although serious concerns have been pointed out that the obtained FRET signals do not simply reflect the change in the expected distance of FRET pairs [58,59]. The trailblazing work for this sensor was reported by Miyawaki et al., in which a genetically encoded calcium indicator composed of two different colored AFP mutants and calmodulin, a Ca²⁺ receptor, has been constructed [36].

The receptor complex that accompanies the dissociation or the association of multiple subunits upon ligand binding was also suitable for the design of FRET-based biosensors (Figure 1b). Zaccoro and coworkers constructed a ratiometric fluorescent biosensor for cyclic adenosine monophosphate (cAMP) on the basis of an intermolecular FRET system between regulatory (R) and catalytic (C) subunit of protein kinase A (PKA) [47,48]. This biosensor can detect the rise of intracellular cAMP concentration by the decrease in the FRET efficiency induced by the dissociation of C subunit from R subunit. More comprehensive information on dual AFP-fused FRET-based biosensors is available in other excellent reviews [60–62].

Recently, Johnsson and coworkers have devised another class of FRET-based semisynthetic biosensors for Zn²⁺ combined with an AFP and a synthetic fluorophore based on the SNAP-tag labeling technology [63]. Because SNAP-tag fusion protein can be covalently modified with O⁶-benzylguanine derivatives in living cells [64], the approach has a potential for in situ preparation of semisynthetic fluorescent biosensors even in living cells.

As an alternative strategy, non-FRET biosensors based on a single circularly permuted (cp) AFP have been developed for targeting Ca²⁺ [38–40,43], cGMP [65], H₂O₂ [66,67], Zn²⁺ [68], and an inositol phosphate derivative [56]. cpAFPs were constructed by connecting original N and C termini by a short peptide linker and regenerating the novel N and C termini at a specific position. A number of cpAFPs with novel termini retained their fluorescence even when a foreign receptor was inserted into the termini [40]. Nakai and coworkers reported a high affinity and signal-to-noise Ca²⁺ indicator (termed as G-CaMP) composed of a single cpGFP and calmodulin [38]. Since G-CaMP responded fast enough to track the dynamic change in Ca²⁺ concentration, this sensor was a powerful tool for visualizing the intracellular dynamics of Ca²⁺ in living cells.

Although the dual AFP-fused FRET-based biosensor is the most facile and robust strategy among AFP-based biosensors, the application of the strategy for FRET-based sensors is essentially difficult in the case of a receptor that undergoes just a slight structural change upon binding to the substrate. In such a case, a sophisticated manipulation of the interplay between the small perturbation of the receptor conformation and the alteration of photochemical property of the AFP chromophore in the ligand-binding event must be required. In the cpAFP-based biosensors, the receptor protein could be placed at the residues near the chromophore that would critically affect the photochemical property of the chromophore in AFP. Morii and coworkers developed a novel cpAFP-based sensor for d-myo-inositol-1,3,4,5-tetrakisphosphate (Ins (1,3,4,5) P₄), from a newly designed split pleckstrin homology (PH) domain of Bruton’s tyrosine kinase (Btk) and a single cpGFP [56] (Figure 2). The resulting split PH domain-cpGFP conjugate, Btk-cpGFP, exhibited bimodal absorption spectra corresponding to the protonated and deprotonated states of the chromophore in GFP. Interestingly, the Btk-cpGFP realized a ratiometric fluorescence detection of Ins (1,3,4,5) P₄ by the excitation of each distinct absorption band, and retained the ligand affinity and the selectivity of the original PH domain.

A major drawback in the above-mentioned two strategies would be the modest fluorescence response upon the ligand binding due to the unavoidable background fluorescence of AFPs. In order to accomplish full suppression of the initial fluorescence, an AFP variant was divided into two fragments. A split GFP displayed quite a low background fluorescence in the separated state and recovered a fluorescence emission significantly by the reassembly of the two fragments when they were attracted in close proximity by a foreign driving force [69]. Based on this split GFP strategy, a receptor composed of two subunits that are associated by binding to the analyte can be converted into a fluorescent biosensor by connecting each of the two subunits with each split GFP fragment. A variety of fluorescent biosensors for detecting Ca²⁺ [43], specific DNA sequences [70,71], DNA methylation [72], and mRNA [73,74], have been constructed. Ghosh and coworkers have devised a split GFP based biosensor for the sequence-specific detection of DNA-methylation at CpG dinucleotides [72]. This sensor consisted of two fragments of the GFP-receptor fusion protein. One GFP fragment was attached by a zinc-finger protein targeting a specific sequence of double stranded DNA, while the other complementary GFP fragment was attached by a methyl-CpG binding domain protein targeting an adjacent methylated CpG dinucleotide site. In the presence of both DNA sites, the reassembly and concomitant fluorescence recovery of the reconstituted GFP was occurred.

In this method, all of the original cysteine residues must be initially substituted with other amino acids followed by the introduction of a unique cysteine residue at the specific position to avoid the non-specific labeling. The incorporation site of a fluorophore could be determined by the three-dimensional structure of the receptor protein obtained from crystallographic analysis.

Hellinga and coworkers have developed a variety of synthetic fluorophore modified biosensors based on E. coli. periplasmic maltose-binding protein (MBP) families [83]. MBPs consist of two domains connected by a hinge region, with a ligand-binding site located at the interface between the two domains, which can adopt two different conformations. The regions induced allosteric conformation changes in response to the ligand binding was identified in MBPs. Environmentally sensitive fluorophores were covalently attached to unique thiols introduced by cysteine mutations within these regions to create biosensors for maltose [84] (Figure 3(a)), glucose [85], ribose [86], and glutamine [86]. Furthermore, extension of the natural diversity by a computational design has provided allosteric fluorescent biosensors for Zn²⁺ [87,88], trinitrotoluene [89], L-lactate [89], serotonin [89], and pinacolyl methyl phosphonic acid [90]. However, since the predicted conformational changes upon substrate binding from the computational design does not necessarily coincide with the conformational changes revealed by the crystal structure, the efficacy of computational design of ligand binding remains controversial [91,92].

Apart from the attachment of fluorophores to allosteric sites of MBP, other group demonstrated that introduction of a fluorophore onto the binding pocket of MBP has also provided a fluorescent biosensor for maltose [94]. The A197C mutant of E. coli. phosphate-binding protein (PBP) was also successfully converted into a fluorescent biosensor for inorganic phosphate (P_i) by the conjugation of thiol-specific coumarin maleimide to Cys197, which is located at the periphery of the binding site [93,95] (Figure 3(b)). The significant fluorescence enhancement of the coumarin-labeled PBP upon binding P_i was discussed with regard to a specific interaction of the coumarin fluorophore with e protein based on the crystallographic analysis [96]. Its rapid response and high selectivity to P_i were suitable for monitoring changes of the P_i concentration in real time [93,97].

Morii and coworkers constructed novel biosensors for inositol 1,4,5-trisphosphate (Ins (1,4,5) P₃) and 1,3,4,5-tetrakisphosphate (Ins (1,3,4,5) P₄) by utilizing the pleckstrin homology (PH) domain of phospholipase C (PLC) δ₁ [98] and the general receptor for phosphoinositides 1 (GRP1) [77] (Figure 4), respectively. In these biosensors, a synthetic fluorophore was attached at the proximity of the ligand-binding site so that the changes in orientation of the fluorophore induced by the substrate binding leads to a sufficient fluorescence response. This structure-based design of synthetic fluorophore-modified biosensors is a powerful method to produce biosensors with high selectivity and appropriate affinity to target inositol derivatives in living cells [77,82,99].

A class A β-lactamase that can efficiently bind to and hydrolyze β-lactam antibiotics has been altered into a fluorescent biosensor for those antibiotics [100,101]. Replacement of the catalytically important residue with a cysteine following a fluorophore-labeling drastically suppressed the enzymatic activity, but instead allowed fluorescence “turn-on” sensing of substrates. Because the biosensor showed a significant fluorescent response to a large variety of β-lactam antibiotics, it can be a promising tool for the discovery of new β-lactamase inhibitors.

Unlike the post-labeling of unique cysteine residues, a mutagenesis technique for direct incorporation of synthetic fluorophores as unnatural amino acids into desired positions in proteins has been developed. Such a site-specific mutagenesis with an expanded genetic code that employed an amber suppression method [102,103] or a four-base codon method [104] in cell-free translation systems has provided a variety of fluorescently modified proteins [105–107].

Hohsaka and coworkers synthesized unnatural amino acids modified with BODIPY derivatives and incorporated two of them into different positions of calmodulin as a donor and acceptor pair for FRET using two four-base codons [107]. The doubly modified calmodulin exhibited a substantial FRET signal in response to the conformational change of calmodulins induced by the addition of the calmodulin binding peptide.

When a three dimensional structure of a receptor protein is not available, it is difficult to convert such a receptor into a fluorescent biosensor by applying the above mentioned methods. An approach enabling a site-specific incorporation of a signal transducer proximal to the binding pocket of intact protein, for which little or no structural information is available, is also highly desirable.

Hamachi and coworkers have reported a new method, which was termed as post-photoaffinity labeling modification (P-PALM) [108], to incorporate a unique benzyl thiol group in the vicinity of the sugar-binding pocket of concavalin A (ConA) that is a well-studied lectin. Incorporation of a thiol group to the ligand-binding pocket of the receptor is conducted by a photoaffinity labeling molecule, which is composed of a ligand moiety, a photoreactive group and a cleavable disulfide group. Subsequently, conjugation of a thiol reactive fluorophore in that site affords a various types of fluorescent biosensors for specific saccharides [109–113]. Recently, the authors have expanded the P-PALM strategy to develop a ligand-directed tosyl (LDT) chemistry-based approach [114,115]. A detailed description of the P-PALM and the LDT strategies is found in other excellent review articles [5,108].

A signaling aptamer covalently modified with a synthetic fluorophore was first explored through a structure-based design by Ellington and co-workers [123]. Similar to the design of protein-based biosensors, a fluorescent reporter was initially placed either in proximity to the ligand-binding site or at the relatively movable position during the ligand-binding event. When acridine or fluoroscein as a signal transducer was introduced close to the binding pocket of anti-ATP RNA or DNA aptamers, respectively, fluorescence emissions of these aptamers were responded to the ATP concentration. The researchers also proposed a direct selection strategy of the signaling aptamers from randomized nucleic acid sequence libraries containing fluorescent nucleotides [124]. A bis-pyrene fluorophore, whose excimer and monomer emission are highly susceptible to the local structural change, was conjugated with an anti-ATP DNA aptamer by Yamana and co-workers [125]. This fluorescent aptamer displayed an efficient ratiometric fluorescence response toward ATP concentration. The electrostatic environment near the 2′-ribose position in nucleic acids highly depends on the local conformational constraint of the nucleotide due to the electrostatic interaction with the adjacent 3′-phosphodiester backbone group [126,127]. Anti-ATP DNA aptamers involving the 2′-pyrene-modified nucleoside devised by the same group served as a fluorescent biosensor with high affinity for ATP [128]. Weeks and coworkers engineered three kinds of RNA aptamer sensors for AMP, tyrosinamide, and argininamide, containing a 2′-BODIPY-modified nucleotides [129]. Furthermore, the researchers demonstrated that this type of sensors could be applicable to the aptamer without any structural information. As another candidate, fluorescent nucleotide analogues such as 2-aminopurine (2AP), 4-amino-6-methylpteridone (6MAP), and 3-methylisoxanthopterin (3MI), were also able to perform efficient signal transduction in aptamer sensors. Katilius and coworkers converted aptamers for human α-thrombin, immunoglobulin E, and platelet-derived growth factor B into the fluorescent biosensors for each targets by utilizing fluorescent nucleotide analogues [130]. The thrombin-binding aptamer with 6MAP showed a 30-fold thrombin-induced fluorescence enhancement.

Analogous to the case of protein-based biosensors, aptamers accompanied by the drastic conformational change upon ligand binding can be altered to biosensors that transduce the ligand binding event into the ratiometric FRET signal by attaching donor and acceptor pairs at the two ends. Takenaka and coworkers employed the formation of a G-quadruplex structure induced by K⁺ as a ratiometric fluorescent sensor for K⁺ [131,132]. By replacement of FRET pairs with a fluorophore and a corresponding quencher, this type of aptamers can be also adapted to the molecular beacon type of signaling method that allows a facile modulation of the fluorescence intensity. Such signaling aptamers known as aptamer beacons are believed to produce a comparatively larger fluorescence signal change than the rational designed aptamer sensors. The aptamer beacon approach has produced a variety of fluorescent biosensors for metal ions [133], small molecules [134–136], and proteins [137–139]. Besides the aptamer beacon composed of single architecture, split aptamer beacons combined with multiple aptamer fragments have also been constructed [140–143]. Li and coworkers reported a structure switching aptamer consisting of three fragments; a template DNA involving aptamer sequence and two small antisense strands bearing a fluorophore or a quencher [142]. This aptamer sensor exploits the target-induced switching between a DNA/DNA duplex and a DNA/target complex. In the absence of the analyte, the emission from the fluorophore was reduced by the quencher because the fluorophore and the quencher were designed that brought close to each other in DNA/DNA duplex form. Addition of the analyte causes the release of the antisense strand with the quencher and the significant increase of fluorescence intensity.

The rational design strategy that has successfully provided fluorescent biosensors including not only the protein-based biosensor but also the aptamer-based biosensor appears to be the most convenient approach for the construction of biosensors. However, this strategy usually requires the redundant optimization of sensor functions because the introduction of fluorophore moieties frequently impairs the original receptor function and does not always guarantee to execute the expected optical signals. Furthermore, since the interplay between the molecular recognition event and the signal-transduction function is essentially unique to the individually constructed biosensor, it is also quite difficult to apply empirically obtained findings from the construction of a biosensor to the other biosensors. A modular strategy that permits facile preparation of biosensors with tailored characteristics by a simple combination of a receptor and a signal transducer has recently emerged as a new paradigm for a versatile design of fluorescent biosensors. Stojanovic and coworkers have reported a modular design of signaling aptamers based on the allosteric regulation of binding events [16]. These chimeric aptamers composed of two modular aptamers, one for the target recognition and another for holding a reporter dye, displayed robust fluorescence responses to three different targets.

Morii and coworkers have recently developed a conceptually new strategy for simultaneous preparation of fluorescent biosensors with diverse functions based on a framework of ribonucleopeptide (RNP) [17] (Figure 5). In the first step to construct the fluorescent RNP sensor, a structure-based design based on the Rev Responsive Element (RRE)-HIV Rev peptide complex affords an RNA-derived RNP library by introducing a randomized nucleotides region as a ligand binding domain adjacent to the RRE segment. In vitro selection method was applied to the RNA-derived RNP library to afford a series of RNP receptors for a given target [144]. In the second step, by the modification of the Rev peptide subunit of the RNP receptor with a fluorophore, the RNP receptor is converted to a fluorescent RNP complex that showed the fluorescent intensity changing upon binding to the target molecule without losing the affinity and the selectivity of parent RNP receptor. The RNP receptors obtained by in vitro selection have a variety of RNA structures and reveal different affinity to the target molecule, and therefore are considered as an RNP receptor library. On the other hand, the peptide subunit is easily converted to a fluorophore-modified Rev peptide library by modifying chromophores with a variety of excitation and emission wavelengths. By taking the advantage of the noncovalent nature of the RNP complex, RNP sensors with desired optical sensing properties can be selected in a high-throughput manner by combining a series of RNA subunits derived from the RNP receptor library and a fluorophore-modified Rev peptide library. Actually, the modular strategy has produced a variety of fluorescent biosensors for targeting ATP [17], GTP [17], histamine [145], phosphotyrosine (pY) [146], and even a peptide fragment containing the pY residue within a defined amino acid sequence [147].

Though the strategy conveniently provides fluorescent RNP sensors, the noncovalent configuration becomes a disadvantage for the practical measurements, for instance, when the sensor concentration is reduced below the nanomolar range where the RNP complex would dissociate to each component. A covalently linking of the RNA and the peptide subunits without sacrificing the sensing function would overcome such disadvantages.